Distribution of Positive COVID-19 Testing Rates

Positive COVID-19 Testing Rates by State

Measure Value
Minimum 0.0259814
Lower Quartile 0.0597063
Median 0.0918429
Upper Quartile 0.1455819
Maximum 0.5446435

Analysis

These three visualizations assess the distribution of positive COVID-19 testing rates. In other words, these visualizations show what proportion of COVID-19 tests have positive results in a way that enables us to compare this rate across states and across regions.

The distribution of COVID-19-positive testing rates by state is positively skewed. The majority of states have low rates of COVID-19-positive tests.

This is evident in the visualizations, but also in the table summary. The median positive testing rate is 9.2%, so half of U.S. states have a positive testing rate of less than 9.2%. The upper quartile positive testing rate is 14.6%, so three-quarters of U.S. states have a positive testing rate of less than 14.6%. The most extreme example of this positive skew are naturally-isolated states like Hawaii and Alaska where less than 2.9% of tests are positive.

However, the maximum testing rate is 55%. The most extreme examples of this are are New York, New Jersey, Oklahoma, and Michigan which have 40%-55% positive testing rates.

A limitation of assessing this metric is that it reflects testing strategy as well as COVID-19 prevalence. A state with a high positive testing rate could reflect one of two scenarios. First, that state could have a low COVID-19 prevalence, but is testing people who are pretty certain to have COVID-19. Second, that state could have a high COVID-19 prevalence, but is testing most people.

For this reason, the distribution of this metric alone is insufficient. In addition, the distribution of prevalence should be considered.


Distribution of COVID-19 prevalence

COVID-19 prevalence by State

Measure Value
Minimum 0.0001522
Lower Quartile 0.0002580
Median 0.0003770
Upper Quartile 0.0006604
Maximum 0.0058677

Analysis

These three visualizations assess the distribution of COVID-19 prevalence. In other words, these visualizations show what proportion of the population tests positive for COVID-19 in a way that enables us to compare this rate across states and across regions.

The distribution of COVID-19-positive testing rates by state is positively skewed. The majority of states have low COVID-19 Prevalence.

This is evident in the visualizations, but also in the table summary. The upper quartile prevalence is 0.07%, so three-quarters of U.S. states have a prevalence of less than 0.07%.

However, the maximum prevalence is 0.59%. The top one-quarter of U.S. states for prevalence include New York, New Jersey, Lousiana, Massachussetts, Connecticut, and Michigan. Each of these states have been reported as being particularly heavily hit by COVID-19, so it is understandable that this is reflected in the data.

Four of the five top states for this metric are in the Northeast. This is reflected in the distributions by region: the prevalence of all Northeastern states is higher than the median prevalence of all non-Northeastern states. This indicates that this region is the hardest-hit by COVID-19, as

A strength of this metric is that it can appropriately compare COVID-19 impapacts by state across states of different population sizes. A limitation of this metric is that it is also somewhat linked to testing strategies: if some states are performing inadequate testing, they will be under-reported in this metric.


COVID-19 cases and industries by employment

Industries with 1% or larger difference in employment between high prevalence and low prevalence states

Industry Top 20 Lowest Prevalence States Top 5 Highest Prevalence States Difference
Manufacturing 6.81% 4.1% 2.71%
Retail Trade 7.51% 5.31% 2.21%
Accommodation Food Services 5.1% 3.34% 1.76%
Health Care Social Assistance 9.16% 7.73% 1.43%
Construction 4.4% 3.16% 1.23%
Agriculture Forestry Fishing Hunting 1.42% 0.25% 1.17%

Analysis

COVID-19 can spread quickly though a workplace. Workplaces differ across industries, and some jobs are more readily transitioned to socially-distanced work than others. For this reason, it is valuable to explore the industries residents of different states are employed in to understand the relationship. In this, we will compare the proportions of industry of employment between the states with the highest COVID-19 prevalence and the states with the lowest COVID-19 prevalence.

The first plot enables the 5 states of highest and lowest prevalence to be compared individually. However, it is different to make conclusions using this plot since these states are very diverse in their environment and economy. 7.18% of jobs in Hawaii are in Accomodations and Food Services compared to only 3.5% in New York, but this is likely an indication that Hawaii’s economy is more tourism-based rather than a predictor of COVID-19 spread.

The second plot enables the 5 states of highest prevalence to be compared against the 20 states of lowest prevalence. Since prevalence distribution is very positively skewed, the 20 states of lowest prevalence should be comparable in the same way as the 5 states of highest prevalence. Using more states limits the influence of outliers.

Considering the second plot and the table highlighting the differences, we find the six industries with the largest difference in employment. It is surprising that nearly all industries (not just those in the table) have a lower share of employment in the top prevalence states compared to the low prevalence states, which is due to the unemployment rate. In future analysis, we should normalize the results to be able to compare the same industry in different states more appropriately. Without that normalization step of making total employment percentages equal to 100% it is challenging to make conclusions on this and it should be explored further. Additionally, since this issue is caused by the unemployment rate, perhaps homelessness is another factor to be considered in our analysis.


Covid-19 and Commuting

Lowest and Highest States by Public Transportation and Covid Cases

Analysis

This graph shows the top 5 and the bottom 5 states as sorted by percent of population that commutes to work using public transportation. All of the states that are in the botom have a rate of close to 0. The states that are in the lowest have less public transportation infrastructure to rely on, so the disease may spread less easily.

Communities and states that have to rely on public transportation may already have higher population density, which makes social distancing harder. Additionally, communities with high rates of public transportation may have a large population of essential workers that rely on that infrastructure. Even during a lockdown, public transportation is needed for the essential, which may aid in the spread of Covid.


Population Ecology and COVID Tests

Knowing how COVID may affect certain demographics can allow certain interventions to be put in place that targets each demographic.

Population Over 60

For instance, with the COVID-19 breakout occurring initially in nursing homes in Washington, it may suggest implementing appropriate infection prevention and control measures for the staff and patients can help prevent the introduction of Covid-19.

COVID-19 is often mentioned to afflict the older population more than the younger ones. Observing the graph above, there seems to be no relation between the percentage of individuals over sixty in a population and the percentage of positive COVID test results, or the number of COVID deaths. This suggests the dataset we are using is likely insufficient; many epidemiological studies such as McMichael et al. (2020) seem to show otherwise.

State density

Observing state population density and positive COVID-19 cases can also be an indicator of how density may affect the spread of COVID-19 given there is a higher chance of contact between agents.

Looking at the graph above, there seems to be no relation between the density of a state and the number of tests that are performed. There also seems to be no obvious relations between the number of tests conducted and the number of positive tests. It might be the case that given the limited testing that has been done, it is difficult to truly tell how many people within the population are infected.


COVID-19 and Housing

## Warning: Use of `apartment_building_df$units_10_plus_proportion` is discouraged.
## Use `units_10_plus_proportion` instead.
## Warning: Use of `apartment_building_df$Covid_Positive_4_4` is discouraged. Use
## `Covid_Positive_4_4` instead.
## Warning: Use of `apartment_building_df$units_10_plus_proportion` is discouraged.
## Use `units_10_plus_proportion` instead.
## Warning: Use of `apartment_building_df$Covid_Positive_4_4` is discouraged. Use
## `Covid_Positive_4_4` instead.
## `geom_smooth()` using method = 'loess' and formula 'y ~ x'

This analysis attempts to explore whether or not states with higher proportions of large apartment buildings (housing 10 or more apartment units or seperate “households”) have seen higher numbers of positive COVID-19 tests. The plot above compares the number of positive COVID-19 tests reported as of April 4, 2020 to the proportion of apartment buildings containing 10 or more units for each state. In order to calculate the proportion of apartment buildings with 10 or more units for each state, the number of buildings reported to have 10+ units was divided by the total number of apartment buildings (ones with 5-9 units + ones with 10+ units). One could hypothesize that states with higher proportions of large apartment buildings may have more dense populations, and therefore more transmission of COVID-19. The plot appears to show a very slight positive correlation between the proportion of large apartment buildings and the number of positive COVID-19 tests, with a jump near 80% large apartment buildings (presumably explained by a few large outliers). For the most part, the correlation between the proportion of large apartment buildings in each state and the number of positive COVID-19 tests seems to be quite weak, and we may need to look for more granular data in order to further explore the theory that apartment building size is proxy to population density and/or COVID transmission.